CN101714255A - Abnormal behavior detection device - Google Patents
Abnormal behavior detection device Download PDFInfo
- Publication number
- CN101714255A CN101714255A CN200910225254A CN200910225254A CN101714255A CN 101714255 A CN101714255 A CN 101714255A CN 200910225254 A CN200910225254 A CN 200910225254A CN 200910225254 A CN200910225254 A CN 200910225254A CN 101714255 A CN101714255 A CN 101714255A
- Authority
- CN
- China
- Prior art keywords
- anomaly
- intensity
- characteristic quantity
- abnormal behaviour
- sniffer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Image Analysis (AREA)
- Alarm Systems (AREA)
Abstract
The invention provides an abnormal behavior detection device which calculates degrees of abnormality of behaviors of a person, an animal, or the like in a video to accurately detect whether an abnormal behavior occurs or not on the basis of degrees of abnormality.The abnormal behavior detection device calculates degrees of abnormality of behaviors in a video for a monitor target acquired by a video acquisition part and determines whether an abnormal behavior occurs or not in accordance with the calculated degrees of abnormality on the basis of a threshold. Since it can be accurately determined whether an abnormal behavior occurs or not on the basis of the degrees of abnormality of behaviors in the video, a quick response to an abnormality can be achieved by giving a warning or informing a guard when the abnormality occurs.
Description
The application divides an application, the application number of its female case application: 200710004061.5, and the applying date: 2007.1.23, denomination of invention: abnormal behaviour sniffer
Technical field
The present invention relates to a kind of abnormal behaviour sniffer that is used to survey the abnormal behaviour of humans and animals etc.
Background technology
For the social labile factors such as growth of corresponding crime incidence, be used to monitor that the quantity that is provided with of the video camera of suspicious figure and suspect vehicle constantly increases.When using numerous video camera like this to monitor,, need to adopt the assistance technology that monitors in order to come effectively monitor area to be monitored with limited supervision Personnel Resources.
As such supervision assistance technology, for example open and disclose a kind of " from three-dimensional data, extracting the method and the device of characteristic quantity " in the 2005-92346 communique the patent of invention spy of Japan of patent documentation 1.The features extraction method of the moving image that is called as the local autocorrelation characteristic of three-dimensional high order is wherein disclosed.And, the method that the behavior that is applied in of this characteristic quantity is discerned and the posture of walking authenticates is also open.
In addition, non-patent literature 1 discloses a kind of method of using the local autocorrelation characteristic of three-dimensional high order to calculate the intensity of anomaly of the personage's behavior in the image.
Patent documentation 1: the patent of invention spy of Japan opens the 2005-92346 communique
Non-patent literature 1: " from a plurality of people's moving image, surveying abnormal operation ", Nan Li Zhuo also, it is opened up, the research report 2004-CVIM-145 of information processing association, on September 11st, 2004 in big Tianjin
Above-mentioned prior art is the technology that a kind of intensity of anomaly of the behavior with personage in the image and animal etc. is calculated as pure (scalar) amount, there is and can not judge the problem in fact whether abnormal behaviour taken place at once.
Summary of the invention
The present invention is based on the problems referred to above and propose, its purpose is to provide a kind of abnormal behaviour sniffer, this abnormal behaviour sniffer can calculate the intensity of anomaly of the behavior of personage in the image and animal etc., and judges definitely according to this intensity of anomaly whether abnormal behaviour has taken place.
In abnormal behaviour sniffer involved in the present invention, have: the image acquisition unit, it obtains the image of monitored object; The intensity of anomaly calculating part, its intensity of anomaly to the image that described image acquisition unit is obtained calculates; And unusual judging part, it is according to threshold value, the intensity of anomaly that calculates from described intensity of anomaly calculating part judges whether to have taken place abnormal behaviour, it is characterized in that, also have make the expression rate of false alarm and not the error curve of the relation between newspaper rate and the described threshold value be presented at display part on the picture.
According to the present invention, can judge definitely whether abnormal behaviour has taken place according to the intensity of anomaly in the image.Therefore, taking place when unusual, can be by giving a warning or, promptly taking corresponding measure to this abnormal behaviour to security personnel's circular.
Description of drawings
Fig. 1 is the block diagram of expression as the functional structure of the abnormal behaviour sniffer of one embodiment of the invention.
Fig. 2 is the process flow diagram of the treatment scheme when representing judgement unusually.
Fig. 3 is the block diagram of the functional structure of expression intensity of anomaly calculating part.
Fig. 4 is the process flow diagram of the flow process of expression intensity of anomaly computing.
Fig. 5 is the key diagram that the frame that uses when the local auto-correlation of three-dimensional high order is calculated is carried out in expression.
Fig. 6 is the key diagram of the local autocorrelative lattice structure of three-dimensional high order (mask pattern).
Fig. 7 is the process flow diagram of flow process of the computing of the local autocorrelation characteristic of the three-dimensional high order of expression.
Fig. 8 is the process flow diagram of the computing flow process of expression transition matrix.
Fig. 9 is the key diagram of the computing of local space.
Figure 10 is the key diagram of the appraisal procedure of intensity of anomaly.
The process flow diagram of the treatment scheme when Figure 11 is the calculating of expression judgment threshold.
Figure 12 is the key diagram of the maximum intensity of anomaly of each scene.
Figure 13 is the rate of false alarm and the key diagram of newspaper rate not.
Figure 14 is the key diagram of error curve.
Figure 15 is the key diagram that the decision of local space is handled.
Figure 16 is the key diagram of the example of monitoring image.
Figure 17 is the process flow diagram of the unusual judgment processing flow process of expression Three Estate.
Figure 18 represents to have the lift appliance of abnormal behaviour sniffer involved in the present invention.
Among the figure: 10-abnormal behaviour sniffer, 20-elevator control gear, 30-camera, 40-lift car, the 50-stern fast, 100-image acquisition unit, 102-intensity of anomaly calculating part, the unusual judging part of 104-, the 106-judgment threshold, 108-judged result, 110-judgment threshold calculating part, 112-communicating department.
Embodiment
Following with reference to accompanying drawing, form of implementation of the present invention is elaborated.
Fig. 1 is the block diagram of expression as the functional structure of the abnormal behaviour sniffer of one embodiment of the invention.This device is made of image acquisition unit 100, intensity of anomaly calculating part 102, unusual judging part 104, judgment threshold calculating part 110 and communicating department 112, and its image according to the monitored object that is obtained by image acquisition unit 100 is surveyed abnormal behaviour.Below describe in regular turn.
Intensity of anomaly calculating part 102 is used to calculate the intensity of anomaly of the image that image acquisition unit 100 obtained.Wherein, so-called intensity of anomaly is a kind of scale, the abnormal behavior degree of mobileses such as personage in its expression image and animal.
The intensity of anomaly that unusual judging part 104 is calculated according to intensity of anomaly calculating part 102 judges whether to have taken place abnormal behaviour, and this result is exported as judged result 108.Use judgment threshold 106 as judgment standard,, be judged as abnormal behaviour does not take place when intensity of anomaly during less than judgment threshold 106.On the contrary, when intensity of anomaly when judgment threshold 106 is above, be judged as abnormal behaviour taken place.
Judgment threshold calculating part 110 is used to calculate unusual judging part 104 when carrying out judgment processing required judgment threshold 106.Wherein, judgment threshold calculating part 110 calculates according to judged result 108, so that the judgement precision of unusual judging part 104 becomes best judgement precision.
This advisory of abnormal behaviour, will take place and give external device (ED) when abnormal behaviour has taken place according to judged result 108 in communicating department 112.The external device (ED) that has notice can be exported alarm with speech form, also can export alarm to monitoring image.And, also can make device such as elevator out of service based on the consideration of secure context.And, can also notify central monitoring position and portable terminal etc. in the mode of telecommunication, take measures to impel it.
Intensity of anomaly calculating part 102, unusual judging part 104, judgment threshold calculating part 110 and communicating department 112 can be realized by arithmetic processing apparatus or the PC of CPU or CPU etc.And judgment threshold and judged result etc. are stored in the memory storages such as semiconductor memory, can read at any time and use in various computings.
Following process flow diagram with reference to Fig. 2, the treatment scheme when judging unusually by the abnormal behaviour sniffer of present embodiment describes.
In step 200,, carry out the processing of step 202 repeatedly, till the user sends END instruction to step 210 with pre-set regulation frequency.
In step 202, by intensity of anomaly calculating part 102, the image that will obtain in image acquisition unit 100 reads in as numerical data.
In step 204,, calculate the intensity of anomaly of the image that in step 202, obtains by intensity of anomaly calculating part 102.
In step 206, by unusual judging part 104, and utilize the intensity of anomaly of in step 204, calculating, judge whether to have taken place abnormal behaviour.
In step 208, the judged result of step 206 is assessed, when being judged as when abnormal behaviour has taken place execution in step 210.
In step 210, by communicating department 112, this advisory of abnormal behaviour will take place given external device (ED).
Following block diagram with reference to Fig. 3 is elaborated to the inner structure of the intensity of anomaly calculating part 102 of Fig. 1.As mentioned above, intensity of anomaly calculating part 102 is calculated the intensity of anomaly of the image that image acquisition unit 100 is obtained as scale, and it is outputed in the unusual judging part 104.This intensity of anomaly calculating part 102 is made of activity extraction unit 300, feature value calculation unit 302, characteristic quantity converter section 304 and intensity of anomaly Rating and Valuation Department 308.Below explanation in regular turn.
Feature value calculation unit 302 is calculated the characteristic quantity of the image that is generated by activity extraction unit 300.When calculating, use the local autocorrelation characteristic (for example, opening the 2005-92346 communique) of known three-dimensional high order with reference to the patent of invention spy of Japan.In the method, the geometric features of the voxel data (voxel data) that will be made up of the image of three continuous frames is calculated as the proper vectors of 251 dimensions.The computing method of relevant this characteristic quantity will explanation in the aftermentioned part.In addition, when calculating the characteristic quantity of image, also can use light stream (optical flow) computing method.So-called optical flow computation method is a kind of tiny area that is conceived to image, and the method with the motion between the frame is calculated as vector for example has detailed argumentation in 243 pages of this works of " Digital Image Processing " (CG-ARTS association).Also whole components of the vector of being calculated by the optical flow computation method can be synthesized characteristic quantity.And, also can be with the average/statistic of disperseing etc. of vector as characteristic quantity.
The feature value vector that characteristic quantity converter section 304 uses 306 pairs of feature value calculation unit 302 of transition matrix to be calculated is carried out linear transformation.By this conversion, extracted the component of the abnormal behaviour that is comprised in the feature value vector.Wherein, be x if establish the feature value vector of calculating by feature value calculation unit 302, transition matrix 306 is A, and the feature value vector after the conversion is x ', and then this conversion can be represented with following formula (1).
x’＝Ax (1)
Intensity of anomaly calculates by being evaluated at the new feature value vector calculated in the characteristic quantity converter section 304 and the irrelevance between the normal data 310 in intensity of anomaly Rating and Valuation Department 308.And result of calculation is output to unusual judging part 104.Wherein, normal data 310 is set of the characteristic quantity of normal behaviour.The computing method of concrete intensity of anomaly will partly illustrate in aftermentioned.
Following process flow diagram with reference to Fig. 4 is elaborated to the intensity of anomaly computing of the step 204 of Fig. 2.
In step 400, from the image that obtains by image acquisition unit 100, extract the part that has produced motion by activity extraction unit 300.
In step 402, calculate the characteristic quantity of the image that in step 400, generates by feature value calculation unit 302.
In step 404, carry out linear transformation by the feature value vector of calculating in 304 pairs of steps 402 of characteristic quantity converter section, to generate new feature value vector.
In step 406, new feature value vector of calculating in step 404 by 308 pairs of intensity of anomaly Rating and Valuation Departments and the irrelevance between the normal data 310 are assessed, to calculate intensity of anomaly.
Following with reference to Fig. 5 to Fig. 7, the characteristic quantity computing of the described moving image of step 402 of Fig. 4 is elaborated.
Fig. 5 is the key diagram of the input data of the local autocorrelation characteristic of above-mentioned three-dimensional high order.The calculating object of characteristic quantity is a moving image, just continuous frame (image) on the time series.In order to calculate the local autocorrelation characteristic of three-dimensional high order, need three width of cloth frames at least.For example, when being given frame number and being the frame 500 of n, this frame and the frame 502 that is positioned at before this frame become the object that characteristic quantity calculates with this three width of cloth frame of frame 504 (corresponding with frame number n-1 and n-2 respectively).
Resolution at the hypothesis frame is vertical h pixel, during horizontal w pixel, by making up three width of cloth frames, can constitute the voxel (cube) of h * w * 3.In the local autocorrelation characteristic computing method of three-dimensional high order, by whole elements, use 3 * 3 * 3 lattice structure 506 in the mode that moves in regular turn at this voxel, extract feature.
In addition, in this enforcement to the situation of three continuous width of cloth frames as process object is described, but also can be with f width of cloth frame arbitrarily as process object.At this moment, be process object with the voxel of h * w * f, calculate the average characteristics amount of the moving image of f width of cloth frame.
The illustration figure of the lattice structure that Fig. 6 uses when being the three-dimensional high order of calculating part autocorrelation characteristic.Lattice structure is used to calculate the correlated characteristic of the part of voxel, and its voxel by 3 * 3 * 3 constitutes.
There are 251 lattice structures in the local autocorrelation characteristic of the three-dimensional high order of bianry image, count, the feature of the input image feature value vector as 251 dimensions can be extracted by the quantity when satisfying each pattern.
Following process flow diagram with reference to Fig. 7 is elaborated to the characteristic quantity computing of the described moving image of step 402 of Fig. 4.
In step 700, feature value vector is carried out initialization.
In step 702, to all voxels as the image of process object, execution in step 704 is to the processing of step 708 repeatedly.That is, as shown in Figure 5,, use lattice structure 506 to scan in regular turn at all voxels of process object.
In step 704, at whole 251 kinds of lattice structures shown in Figure 6, execution in step 706 is to the processing of step 708 repeatedly.
In step 706, judge whether the pixel corresponding with the lattice structure of process object all is 1.If judged result is sure, then execution in step 708.
In step 708, the lattice structure characteristic of correspondence amount vector components with process object is only added 1.
By above-mentioned a series of processing, can calculate the feature value vector of 251 related dimensions of the local auto-correlation of three-dimensional high order.
Following process flow diagram with reference to Fig. 8 describes the computation sequence of the transition matrix 306 of Fig. 3.
In step 800, at the image of the more than one normal scene in the memory storage that is stored in semiconductor memory etc. for study in advance repeatedly execution in step 400 to the processing of step 402.
In step 400, as shown in Figure 4, from the image that generates by image acquisition unit 100, extract the part that has produced motion by activity extraction unit 300.
In step 402, as shown in Figure 4, calculate the characteristic quantity of the image that in step 400, generates by feature value calculation unit 302.
In step 802,, carry out principal component analysis at the set that calculates to the characteristic quantity of normal scene.Principal component analysis is a kind of in the multivariate analytic method.It generates the compositional variable that is called as principal component by according to several variablees in irrelevant each other mode, can conclude the information that a plurality of variable has.This principal component analysis is the method for often using in the parsing of multivariate data, owing to for example in " multivariate that is easily understood is resolved " (Tokyo bibliogony) this works detailed explanation is arranged, therefore, omits its detailed description at this.Carry out principal component analysis by set, can obtain 251 major components and eigenwert 251 feature value vector of tieing up.
In step 804,, calculate the local space low to the contribution rate of normal behaviour according to the result of the principal component analysis in the step 802.Transition matrix 306 is set to the matrix that is used for feature value vector is converted to the vector of this local space.
Following with reference to Fig. 9, the computing of the described local space of step 804 of Fig. 8 is elaborated.Fig. 9 is illustrated in the accumulation contribution rate of each principal component that obtains in the principal component analysis shown in the step 802.So-called accumulation contribution rate is to obtain by with from big to small order the contribution rate of each principal component being added, it is a kind of index, the size of the explanation degree that expression principal component is before this made the original quantity of information that has of the data of analytic target.For example, if the accumulation contribution rate 900 till the 3rd principal component is 90%, represent that then the 1st principal component to the 3 principal components have expressed 90% of the original quantity of information of these data.On the other hand, the quantity of information that had of remaining the 4th principal component to the 251 principal components then only accounts for 10% of the original quantity of information of data.
As can be seen, the local space that is made of the 1st principal component to the 3 principal components is big to the contribution rate of normal behaviour from above-mentioned explanation.And the local space that is made of the 4th principal component to the 251 principal components is little to the contribution rate of normal behaviour.
Like this, by to accumulate contribution rate, can obtain the local space little to the contribution rate of normal behaviour as judgment standard.
Following with reference to Figure 10, the method for the evaluation process of the described intensity of anomaly of step 406 of Fig. 4 is described.Being evaluated in the little local space of the contribution rate to normal behaviour shown in Figure 9 of intensity of anomaly carried out.This be because, in this local space, the dispersion of the characteristic quantity of normal behaviour is little, and the behavior beyond the normal behaviour, promptly during abnormal behaviour, this divides breaking up to become big.
Below, this local space is assumed to the space that is made of the principal component below the n principal component.Original this arrangement space should be the local space of 252-n dimension, but for convenience of explanation, has represented the principal component that contribution rate is big, i.e. n principal component and n+1 principal component with the form of two axles among Figure 10.The set 1000 of characteristic quantity is the set of normal data 310.In to the little local space of the contribution rate of normal action, characteristic quantity with the set center of gravity xn1004 be the center, be distributed in this center around near.Therefore, if the feature value vector x1002 of the current image of assessing near the set 1000 of characteristic quantity, can judge then that it is normal, and if a good distance off can judge that then it is unusual.Wherein, distance 1006 between the two is used as intensity of anomaly.
Can adopt the Euclidean distance computing method that assesses the cost low to come distance between the set 1000 of calculated characteristics amount vector x 1002 and characteristic quantity.But in the present embodiment, adopted the mahalanobis distance computing method of the dispersion of characteristic quantity set having been made consideration.The reverse matrix of variance-covariance matrix of supposing the set 1000 of characteristic quantity is S
-1, then can calculate mahalanobis distance by following formula (2).
D
2＝(x-x
n)
tS
-1(x-x
n) (2)
Following process flow diagram with reference to Figure 11, the treatment scheme when the abnormal behaviour sniffer by present embodiment is calculated judgment threshold 106 describes.
In step 1100, by intensity of anomaly calculating part 102, to all assessments with the scene processing of execution in step 1102 repeatedly.So-called assessment is the image storehouse of normal scene and unusual scene with scene, and this assessment is with being endowed the result that this device should be judged on the scene.
In step 1102, the assessment that calculates process object by judgment threshold calculating part 110 is with maximum in the intensity of anomaly of each frame of scene, as the typical value of the intensity of anomaly of this scene.
In step 1104,, and, generate error curve described later according to the maximum intensity of anomaly of each scene of in step 1102, calculating by judgment threshold calculating part 110.So-called error curve is expression rate of false alarm and the curve that how to change according to judgment threshold of newspaper rate not, and wherein, rate of false alarm is the ratio that normal scene is judged as mistakenly unusual scene, and the newspaper rate is not the ratio that unusual scene is judged as mistakenly normal scene.
In step 1106, according to the error curve that in step 1104, generates, the decision suitable threshold.This determining method describes in detail in the aftermentioned part.
Following with reference to Figure 12, the computation sequence of the maximum intensity of anomaly of described each scene of step 1102 of Figure 11 is elaborated.Transverse axis in the curve map is represented frame number, and the longitudinal axis is represented intensity of anomaly.The figure shows the situation of change of the intensity of anomaly of each width of cloth frame when assessment assessed in regular turn with scene.Curve 1200 is the assessment results to the intensity of anomaly of each width of cloth frame of scene 1.Among the figure, the maximal value of intensity of anomaly is point 1204, with the maximum intensity of anomaly of this value as scene 1.Equally, curve 1202 is the assessment results to the intensity of anomaly of scene 2, with the maximum intensity of anomaly of point 1206 corresponding intensity of anomalys as scene 2.
Following with reference to Figure 13 and Figure 14, the calculation procedure of the described error curve of step 1104 of Figure 11 is elaborated.
Figure 13 is the bar shaped curve map that is illustrated in the maximum intensity of anomaly of each scene of calculating in the step 1102 of Figure 11.In the present embodiment, scene 1 to scene 7 is normal scene, and scene 8 to scene 14 is unusual scene.Can know that from this chart the maximum intensity of anomaly with normal scene is little, and the big tendency of maximum intensity of anomaly of unusual scene.
Wherein, as unusual and normal judgment standard, introduced judgment threshold 1300.When maximum intensity of anomaly during less than this judgment threshold 1300, this scene can be regarded as normal scene, and when maximum intensity of anomaly when this judgment threshold 1300 is above, this scene can be regarded as unusual scene.
The abnormality detection performance of this device can adopt rate of false alarm and not the newspaper rate assess.So-called rate of false alarm is meant the ratio that normal scene is judged as mistakenly unusual scene, and this value is the smaller the better.Under the situation of Figure 13, in normal 7 scenes, the maximum intensity of anomaly 1302 of scene 5 has surpassed judgment threshold 1300, is judged mistakenly as unusual its.At this moment, rate of false alarm is 1/7=14%.And the newspaper rate is not meant the ratio that unusual scene is judged as mistakenly normal scene, and this value is the smaller the better.Under the situation of Figure 13, in 7 unusual scenes, because the maximum intensity of anomaly 1304 of scene 12 is less than judgment threshold 1300, so it is judged mistakenly as normal scene.At this moment, the newspaper rate is not 1/7=14%.
Rate of false alarm and not the newspaper rate change with the value of judgment threshold.Its situation of change as shown in figure 14.Among the figure, curve 1400 expression rate of false alarms, curve 1402 is represented not newspaper rate.As can be seen from the figure, rate of false alarm and be not this those long relations that disappear between the newspaper rate.That is, if set bigger judgment threshold to reduce rate of false alarm, then the newspaper rate does not increase.On the contrary, if set less judgment threshold to reduce not newspaper rate, then rate of false alarm increases.In the step 1106 of Figure 11, set judgment threshold 106 according to this curve map, to realize detection performance even more ideal concerning the user of this device.
As the candidate of judgment threshold 106, rate of false alarm and the point 1404 that equates of newspaper rate are not for example arranged, promptly equate error rate (Equal Error Rate, EER).At this moment, judgment threshold 1406 that will be corresponding with equal error rate is set at judgment threshold 106.
In addition, also can be that the judgment threshold 1410 that 0 o'clock judgment threshold 1408 or rate of false alarm are at 0 o'clock is set at judgment threshold with newspaper rate not.
And, because the alternative condition of these judgment thresholds is different with service condition because of the application target of this device, so, also can be arranged to select by the user of this device.
In addition, can the error curve of Figure 14 be pointed out to the user by output units such as displays.At this moment, also can be arranged to and on picture, to select judgment threshold by input medias such as mouse or keyboards.
By said method, have and to improve the effect that degree of freedom is set.
By the above embodiments, can judge whether automatically to have taken place unusually according to the intensity of anomaly in the image.Thus, when unusual the generation, by sending alarm or, can taking measures unusually to this immediately to security personnel's circular.And, can determine judgment threshold automatically according to the assessment result of the scene of assessing usefulness, thereby obtain the judgment threshold of the best as unusual generation judgment standard.Therefore, even taken place at the unusual object of judging also only to need to generate the scene of assessment usefulness under the situation of change, just can take appropriate measures neatly.
In the embodiment of above explanation, the less local space of contribution rate to the proper space shown in Figure 9 decides according to predefined accumulation contribution rate, but can not calculate local space yet, and can adopt the mode that makes detection accuracy become the best to decide local space according to fixing accumulation contribution rate.
Followingly the determining method of above-mentioned local space is described with reference to Figure 15.EER curve 1500 among Figure 15 is illustrated in the local space that is made of 251 principal components of a n principal component to the EER value when calculating intensity of anomaly and the relation between the n.It is generally acknowledged that EER illustrated in fig. 14 is more little, detection performance is high more.Therefore, only needing to select the principal component corresponding with the minimum value 1502 of EER curve 1500 to count n gets final product.
Thus, can adopt and make detection accuracy become best mode to determine local space automatically.Therefore, the user need not require efforts, and just can realize best detection performance.
Figure 16 has been to use the illustration figure of the monitoring image of abnormal behaviour sniffer of the present invention.This monitoring image is presented on the display device of the PC that is built-in with the abnormal behaviour sniffer or monitoring terminal.Image in this illustration is the image of being taken by the camera in the lift car.
In addition, also can not divide into normal and unusual these two grades, and the grade of dividing into more than three kinds is come the display abnormality degree.Zone 1608 illustrations come the situation of display abnormality degree with Three Estate.For example, by representing with blueness that normally yellow is represented mile abnormality, red expression severe is unusual, and the supervision personnel can grasp situation intuitively.
Following process flow diagram with reference to Figure 17 is elaborated to the unusual judgment processing of a plurality of grades shown in the zone 1608 of Figure 16.At this, till calculating from certain time point of past to current point in time during be judged as the shared time ratio of unusual state, and carry out the judgement of a plurality of grades according to this ratio.
In step 1700, till calculating from certain time point of past to current point in time during in be judged as the shared ratio p of unusual state.
In step 1702, relatively whether this occupation rate p is less than the preassigned blue threshold value pb that judges usefulness.If result relatively is for being that then execution in step 1704.If whether comparative result, then execution in step 1706 is to step 1710.
In step 1704, the judged result decision is blue back end process.
In step 1706, relatively whether occupation rate p is less than the preassigned yellow threshold value py that judges usefulness.If result relatively is for being that then execution in step 1708.If whether comparative result, then execution in step 1710.
In step 1708, the judged result decision is yellow back end process.
In step 1710, the judged result decision is red back end process.
In above-mentioned processing, minute Three Estate is shown the situation of judged result is described, but also can adopt more grade to represent judged result.At this moment, only need the threshold value of pre-set each grade to get final product.So, by representing judged result, has the effect of the method for monitoring that can provide convenience for the overseer with a plurality of grades.
According to the foregoing description, can grasp the corresponding relation between the intensity of anomaly of the presentation content of evaluation object and image easily.Therefore, even wrong report for example occurred, the overseer also can promptly confirm whether really to have taken place abnormal behaviour by image at an easy rate.
Figure 18 represents to have the lift appliance of abnormal behaviour sniffer involved in the present invention.Wherein, in Figure 18, omitted the diagram of the driving mechanism of windlass, hoist cable and damper weight etc.Obtain image in the lift car by being arranged on cameras 30 in the lift car 40, and by stern fast 50 this signal of video signal is sent to and is arranged in the hoist trunk or abnormal behaviour sniffer 10 that hoist trunk is outer.When having detected in the image in lift car 40 when unusual, abnormal behaviour sniffer 10 will be represented that unusual judging result signal outputs to and be arranged in the hoist trunk or the elevator control gear 20 of Machine Room.After elevator control gear 20 received the unusual judging result signal of expression, the control windlass was controlled the door that elevator door driving device is opened lift car 40 and elevator lobby simultaneously so that lift car 40 rests on the nearest floor.Perhaps control device 20 makes alarm devices such as the alarm action that is arranged in the lift car 40.According to above-mentioned lift facility, can make the personnel that ride in the same lift car abnormal behavior person that speeds away, perhaps suppress the abnormal behaviour in the lift car, thus the security that can improve lift facility.In addition, abnormal behaviour sniffer 10 also can be arranged in the lift car 40.At this moment, judging result signal sends control device 20 to via stern fast 50.
Claims (9)
1. abnormal behaviour sniffer has:
The image acquisition unit, it obtains the image of monitored object;
The intensity of anomaly calculating part, its intensity of anomaly to the image that described image acquisition unit is obtained calculates; And
Unusual judging part, it is according to threshold value, and the intensity of anomaly that calculates from described intensity of anomaly calculating part judges whether to have taken place abnormal behaviour,
In the described abnormal behaviour sniffer, also have make the expression rate of false alarm and not the error curve of the relation between newspaper rate and the described threshold value be presented at display part on the picture.
2. abnormal behaviour sniffer as claimed in claim 1 is characterized in that,
Also has the input mechanism of on the picture that shows described error curve, selecting described threshold value.
3. abnormal behaviour sniffer as claimed in claim 1 is characterized in that,
Described intensity of anomaly calculating part has:
The activity extraction unit, it extracts the part that has produced motion from described image;
Feature value calculation unit, it calculates first characteristic quantity of the image that is generated by described activity extraction unit;
The characteristic quantity converter section, it is converted to second characteristic quantity in the linear transformation mode with described first characteristic quantity; And
Intensity of anomaly Rating and Valuation Department, it compares described second characteristic quantity and pairing the 3rd characteristic quantity of normal behaviour and calculates intensity of anomaly.
4. abnormal behaviour sniffer as claimed in claim 3 is characterized in that,
Described feature value calculation unit adopts the local autocorrelation characteristic of three-dimensional high order to calculate described first characteristic quantity.
5. abnormal behaviour sniffer as claimed in claim 3 is characterized in that,
Intensity of anomaly calculates according to the Euclidean distance between the center of gravity of the set of described second characteristic quantity and described the 3rd characteristic quantity in described intensity of anomaly Rating and Valuation Department.
6. abnormal behaviour sniffer as claimed in claim 3 is characterized in that,
Intensity of anomaly calculates according to the mahalanobis distance of the set of described second characteristic quantity and described the 3rd characteristic quantity in described intensity of anomaly Rating and Valuation Department.
7. as any described abnormal behaviour sniffer of claim 1 to 6, it is characterized in that,
Have the judgment threshold calculating part, this judgment threshold calculating part is set described unusual judging part required described threshold value when carrying out judgment processing in the mode that can obtain best judged result.
8. abnormal behaviour sniffer as claimed in claim 7 is characterized in that,
The setting of described judgment threshold calculating part and rate of false alarm and any one corresponding threshold in the equal error rate that equates of newspaper rate, 0% rate of false alarm and 0% the not newspaper rate not.
9. as any described abnormal behaviour sniffer of claim 3 to 6, it is characterized in that,
Described characteristic quantity converter section, with rate of false alarm and the mode of the equal error rate minimum that equates of newspaper rate generate the transition matrix of described linear conversion.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2006040111A JP4701100B2 (en) | 2006-02-17 | 2006-02-17 | Abnormal behavior detection device |
| JP2006-040111 | 2006-02-17 |
Related Parent Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2007100040615A Division CN101025825B (en) | 2006-02-17 | 2007-01-23 | Abnormal action detecting device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101714255A true CN101714255A (en) | 2010-05-26 |
| CN101714255B CN101714255B (en) | 2012-08-22 |
Family
ID=38497094
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2009102252542A Active CN101714255B (en) | 2006-02-17 | 2007-01-23 | Abnormal behavior detection device |
| CN2007100040615A Active CN101025825B (en) | 2006-02-17 | 2007-01-23 | Abnormal action detecting device |
Family Applications After (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2007100040615A Active CN101025825B (en) | 2006-02-17 | 2007-01-23 | Abnormal action detecting device |
Country Status (3)
| Country | Link |
|---|---|
| JP (1) | JP4701100B2 (en) |
| CN (2) | CN101714255B (en) |
| HK (2) | HK1107861A1 (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI614698B (en) * | 2014-10-23 | 2018-02-11 | 美和學校財團法人美和科技大學 | Detection system for estrus of quadruped |
| CN111149129A (en) * | 2017-09-21 | 2020-05-12 | 株式会社安德科技术 | Abnormality detection device and abnormality detection method |
| CN113348420A (en) * | 2019-03-28 | 2021-09-03 | 三菱动力株式会社 | Plant monitoring device, plant monitoring method, and program |
| CN116453247A (en) * | 2023-06-13 | 2023-07-18 | 三峡高科信息技术有限责任公司 | Intelligent lock control system based on Internet of things technology |
Families Citing this family (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP4842197B2 (en) * | 2007-04-17 | 2011-12-21 | 財団法人ソフトピアジャパン | Abnormal operation detection device using multiple divided images, abnormal operation detection method, and abnormal operation detection program |
| JP4925120B2 (en) * | 2007-07-02 | 2012-04-25 | 独立行政法人産業技術総合研究所 | Object recognition apparatus and object recognition method |
| JP5334008B2 (en) * | 2007-10-31 | 2013-11-06 | 東芝エレベータ株式会社 | Abnormal operation detection device |
| JP4654347B2 (en) * | 2007-12-06 | 2011-03-16 | 株式会社融合技術研究所 | Abnormal operation monitoring device |
| JP4663756B2 (en) * | 2008-04-28 | 2011-04-06 | 株式会社日立製作所 | Abnormal behavior detection device |
| JP4663767B2 (en) * | 2008-09-01 | 2011-04-06 | 株式会社日立製作所 | Image surveillance system |
| JP2010117267A (en) * | 2008-11-13 | 2010-05-27 | Nippon Telegr & Teleph Corp <Ntt> | Abnormal state detecting system, and method and program of selecting sensor therefor |
| US8165349B2 (en) * | 2008-11-29 | 2012-04-24 | International Business Machines Corporation | Analyzing repetitive sequential events |
| JP2011048547A (en) * | 2009-08-26 | 2011-03-10 | Toshiba Corp | Abnormal-behavior detecting device, monitoring system, and abnormal-behavior detecting method |
| JP5155279B2 (en) * | 2009-10-29 | 2013-03-06 | 株式会社日立製作所 | Centralized monitoring system and centralized monitoring method using multiple surveillance cameras |
| JP2012165240A (en) * | 2011-02-08 | 2012-08-30 | Sony Corp | Moving image processing apparatus, moving image processing method, and program |
| CN102829967B (en) * | 2012-08-27 | 2015-12-16 | 中国舰船研究设计中心 | A kind of time domain fault recognition method based on regression model index variation |
| CN103164993B (en) * | 2013-02-22 | 2015-02-11 | 福建华映显示科技有限公司 | Digital teaching system and screen monitoring method thereof |
| JP5635637B2 (en) * | 2013-03-15 | 2014-12-03 | ヤフー株式会社 | Animal abnormality detection device, animal abnormality detection method, and program |
| KR101441107B1 (en) | 2013-04-29 | 2014-09-23 | 주식회사 에스원 | Method and apparatus for determining abnormal behavior |
| CN105405150B (en) * | 2015-10-21 | 2019-04-30 | 东方网力科技股份有限公司 | Anomaly detection method and device based on fusion feature |
| JP6942472B2 (en) | 2017-01-13 | 2021-09-29 | キヤノン株式会社 | Video recognition device, video recognition method and program |
| EP3406556A1 (en) * | 2017-05-23 | 2018-11-28 | Otis Elevator Company | Elevator doorway display systems for elevator cars |
| CN110407052B (en) * | 2019-08-02 | 2020-11-10 | 浙江新再灵科技股份有限公司 | Method and system for detecting violent movement behaviors in elevator |
| JP7433193B2 (en) | 2020-11-17 | 2024-02-19 | 三菱電機株式会社 | Suspicious behavior monitoring support system and suspicious behavior monitoring support method |
| CN112633222B (en) * | 2020-12-30 | 2023-04-28 | 民航成都电子技术有限责任公司 | Gait recognition method, device, equipment and medium based on countermeasure network |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS63253478A (en) * | 1987-04-10 | 1988-10-20 | Hitachi Ltd | Method and device for monitoring of fish picture |
| JP2001021309A (en) * | 1999-07-12 | 2001-01-26 | Toshiba Tec Corp | Individual body authentication method and individual person authentication method |
| JP2001216585A (en) * | 2000-02-03 | 2001-08-10 | Sekisui Chem Co Ltd | Abnormal behavior judging system |
| JP2004152087A (en) * | 2002-10-31 | 2004-05-27 | Fuji Photo Film Co Ltd | Method and apparatus for extracting feature vector of image |
| JP4368767B2 (en) * | 2004-09-08 | 2009-11-18 | 独立行政法人産業技術総合研究所 | Abnormal operation detection device and abnormal operation detection method |
| CN1617512A (en) * | 2004-11-25 | 2005-05-18 | 中国科学院计算技术研究所 | Adaptive network flow forecasting and abnormal alarming method |
-
2006
- 2006-02-17 JP JP2006040111A patent/JP4701100B2/en active Active
-
2007
- 2007-01-23 CN CN2009102252542A patent/CN101714255B/en active Active
- 2007-01-23 CN CN2007100040615A patent/CN101025825B/en active Active
-
2008
- 2008-01-30 HK HK08101176.6A patent/HK1107861A1/en not_active IP Right Cessation
-
2010
- 2010-09-13 HK HK10108671.7A patent/HK1142156A1/en not_active IP Right Cessation
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI614698B (en) * | 2014-10-23 | 2018-02-11 | 美和學校財團法人美和科技大學 | Detection system for estrus of quadruped |
| CN111149129A (en) * | 2017-09-21 | 2020-05-12 | 株式会社安德科技术 | Abnormality detection device and abnormality detection method |
| CN113348420A (en) * | 2019-03-28 | 2021-09-03 | 三菱动力株式会社 | Plant monitoring device, plant monitoring method, and program |
| CN116453247A (en) * | 2023-06-13 | 2023-07-18 | 三峡高科信息技术有限责任公司 | Intelligent lock control system based on Internet of things technology |
| CN116453247B (en) * | 2023-06-13 | 2023-09-19 | 三峡高科信息技术有限责任公司 | Intelligent lock control system based on Internet of Things technology |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101025825A (en) | 2007-08-29 |
| CN101714255B (en) | 2012-08-22 |
| HK1142156A1 (en) | 2010-11-26 |
| CN101025825B (en) | 2010-08-04 |
| HK1107861A1 (en) | 2008-04-18 |
| JP4701100B2 (en) | 2011-06-15 |
| JP2007219865A (en) | 2007-08-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101025825B (en) | Abnormal action detecting device | |
| JP4663756B2 (en) | Abnormal behavior detection device | |
| US8619135B2 (en) | Detection of abnormal behaviour in video objects | |
| JP4369961B2 (en) | Abnormality detection device and abnormality detection program | |
| JP2018101317A (en) | Abnormality monitoring system | |
| JP4718253B2 (en) | Image abnormality detection device for surveillance camera | |
| KR102058452B1 (en) | IoT Convergence Intelligent Video Analysis Platform System | |
| CN109219956B (en) | Monitoring device | |
| KR102182008B1 (en) | Control system and method using integrated environmental monitoring | |
| KR101454644B1 (en) | Loitering Detection Using a Pedestrian Tracker | |
| US20090322875A1 (en) | Surveillance system, surveillance method and computer readable medium | |
| KR20160093253A (en) | Video based abnormal flow detection method and system | |
| JP2007018324A (en) | Apparatus, method and program for detecting intruding object by image processing | |
| JP6469139B2 (en) | Information processing apparatus, information processing method, and program | |
| US12046078B2 (en) | Accident detection device and accident detection method | |
| CN117422991A (en) | Intelligent mine detection system and method based on big data and readable storage medium | |
| CN115346170B (en) | Intelligent monitoring method and device for gas facility area | |
| KR102464196B1 (en) | Big data-based video surveillance system | |
| JP2005284652A (en) | Video monitoring method and apparatus using moving vector | |
| JP2008052683A (en) | Suspicious individual determination system and profile distribution system | |
| KR102286229B1 (en) | A feature vector-based fight event recognition method | |
| JP5228067B2 (en) | Abnormal behavior detection device | |
| JP2021087031A (en) | Information processing device, information processing method, monitoring system, and program | |
| KR101516672B1 (en) | Method for detecting video that utilize smarter video control center using the same system | |
| WO2020139071A1 (en) | System and method for detecting aggressive behaviour activity |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 1142156 Country of ref document: HK |
|
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: GR Ref document number: 1142156 Country of ref document: HK |